@infoinlet/mcp-jp-doc-intel

MCP server for Japanese document intelligence — a thin client over the hosted jp-doc-intel API. Point your agent at it to extract structured data from invoices, business cards, and scanned/handwritten documents.

Free to install; calls a paid hosted API. Start a free trial / get a key at infoinlet.com. No model keys needed — the server is just a client; all OCR/VLM/LLM work runs on the hosted service.

Tools

| Tool | What it does | |---|---| | extract_invoice | Vendor, invoice no., validated 適格請求書 T-number, dates (元号→Gregorian), subtotal/tax/total (zero-decimal aware), tax breakdown, line items. | | extract_document | OCR/parse any document to text; optional JP document-kind classification. Multi-page PDFs handled page-by-page. | | extract_handwriting | VLM OCR for handwriting, hanko (stamps), faxed/scanned artifacts. | | extract_meishi | Business-card → name, kana, company, title, email, phone, address. | | classify_text | Cheaply classify pasted text into a JP document kind. | | usage | Your key's usage/quota (trial: used/limit/remaining/expiry; subscribed: unlimited). |

File tools accept either a local path or content_base64 + filename.

Configure

JP_DOC_INTEL_API_KEY=sk-...                       # required
JP_DOC_INTEL_BASE_URL=https://api-docintel.knowlez.com  # optional override

Claude Desktop / any MCP client

{
  "mcpServers": {
    "jp-doc-intel": {
      "command": "npx",
      "args": ["-y", "@infoinlet/mcp-jp-doc-intel"],
      "env": { "JP_DOC_INTEL_API_KEY": "sk-..." }
    }
  }
}

Notes

• Errors are surfaced plainly: 401 (bad/missing key), 402 (trial quota used up — subscribe), 429 (rate limited).
• Read-only against your account; the only state it changes is consuming your quota.